This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: [Patch,libgomp] Fix race in gomp_mutex_unlock.

From: David Daney <ddaney at caviumnetworks dot com>
To: Jakub Jelinek <jakub at redhat dot com>
Cc: GCC Patches <gcc-patches at gcc dot gnu dot org>, Adam Nemet <anemet at caviumnetworks dot com>
Date: Mon, 03 Aug 2009 15:36:29 -0700
Subject: Re: [Patch,libgomp] Fix race in gomp_mutex_unlock.
References: <4A775F2E.5030200@caviumnetworks.com> <20090803222555.GC4462@tyan-ft48-01.lab.bos.redhat.com>

Jakub Jelinek wrote:

On Mon, Aug 03, 2009 at 03:05:34PM -0700, David Daney wrote:
OK to commit?
No.  Instead
#include "config/linux/ia64/mutex.h"
in config/linux/mips/mutex.h.

That doesn't make any sense to me.

On machines where the __sync_synchronize () is unneeded, it will expand to nothing. So for things like x86 my patch shouldn't change the code in any way.

But instead you want a default implementation that is broken according to GCC's own atomic memory operations documentation, and to maintain separate correct implementations for machines with weak ordering.

Can you expound on why your suggestion is better than a combination of my patch *and* removing config/linux/ia64/mutex.h?

David Daney

Follow-Ups:
- Re: [Patch,libgomp] Fix race in gomp_mutex_unlock.
  - From: Richard Henderson

References:
- [Patch,libgomp] Fix race in gomp_mutex_unlock.
  - From: David Daney
- Re: [Patch,libgomp] Fix race in gomp_mutex_unlock.
  - From: Jakub Jelinek

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]